As Intel on 14 November 2002.

… the Hyper-Threading technology with the Pentium 4 3.06 GHz introduced us it was clear that a feature which several tasks favoured the parallel processing, can be tested not conventionally.

In addition to a traditional benchmarking, we published an article who has tried a plethora of benchmarks to determine the benefits of HT technology in parallel processing of two programs on the same day: “Intel Pentium 4 in the Hyper-Threading special”, the article was packaged in the archive or games played, while SETI@home version, the media player, flask or the McAfee virus scanner was running in the background. Not just practical but it fulfilled the purpose very effectively.

A uniform picture withdrew over the entire course of the benchmark: while the processor with no Hyper-Threading under parallel load compared to processing without additional hurdle broke a. With Hyper-Threading, this performance loss turned out however noticeably less. The problem: The setting up of a such lab, which is often subject of chance, which has nothing to find chance in meaningful benchmarks, however, requires a significant overhead to the clinically patch test environment in the program after program download is recorded. Here, the direct comparison is given without a doubt. But what is compared?

In the last two years we limited always more short-term deployment of test products on the part of the manufacturer us thus after our first foray into the world of the parallel benchmark harness again to the serial tests, not least effectively prevented a return to larger test methods. Even if we ourselves in the truest sense of the word actually always again weird occurred when before benchmarks the system was rebooted, annoying system tools have been disabled and one on his private machine a few minutes later from the flagship overloaded with programs with ICQ, MSN, Outlook, mIRC, virus scanner and Firewall started a game, forgot, or we ultimately again rejected our findings at that time – the time allowed to any other test method.

And so we moved test to test the conclusion that AMD’s CPUs in games are currently unbeaten at the top and landed itself in applications, an eternal Pentium parade discipline, partially before the blue giant went with virtually every Web and printed globally compliant.

The members of the HardTecs4U, where this approach was also contrary were it now, who had the final idea, itself made the effort, the time taken and chased selected AMD and Intel processors by a benchmark course, the real conditions of use, found that a large number of readers and customers should be tried, the lightly. In addition to the actual benchmarks Norton AntiVirus 2003, ICQ Lite, ZoneAlarm, Firefox and Outlook Express on the test system ran so mIRC. In addition, it has replaced the usual 512 resolution with the nearest equivalent with 1280 and also also AF and FSAA back resorted, as it is today no longer a problem with current high end cards. That many players emulate still the frame rates in the digit, you could keep this decision contrary to.

At first glance, the results look pretty good and one is tempted to do so to sign the bottom line that currently probably but an Intel CPU is the better desktop Porzessor, HardTecs4U of. In particular Hyper-Threading were the Pentium 4 in the workspace to a victory over the AMD counterpart. Ultimately, this evaluation is then however – as so often – for the point of view from which you look at the results.

Can Intel save themselves a narrow victory ago AMD in the course of the HardTecs4U in certain games and therefore the results of our clinical trials with 1024 x 768 provide completely on its head. Ultimately, however, all tested CPUs are within less frames-per-second. The decisive factor for the stalemate the high graphical splendour seems away from the background applications much more, which must conjure up the graphics card with effort and hardship on the screen and to the neck of the bottle is clear. Practice? Here, there no. How the Intel CPUs with hyperthreading in lower resolutions would have cut off, was not considered unfortunately.

In the following application tests Intel can dominate then almost everywhere significantly the competition part and the 570J beats the 4000 + just as clearly as the 530 the 3000 +. But if one takes a look at our last performance rating, was this – if not quite so clear – in a clinical environment also the case. One advantage thanks to Hyper-Threading is not to assign, on the head placed were old evaluations but not and that the P4 felt rounder ran, can be – oh Graus – so not in a graph hold.

At the end, only the vertrakte matter how to performance of the processor now defines the practice-oriented and created a benchmark course that meets this definition remains. “Should you slow down the processors in gaming through high resolutions, because the average user runs these settings and performance described then as equivalent?”, a more specific question should be. Should one separate performance and value this performance at the end? True masses on operands of different resolutions and in the end also graphics cards, memory constellations and motherboards (especially for processors that have no integrated memory controller, varies the speed from chipset to chipset) would have to be used for such a purpose. You will notice the perfect, practical test seems impossible, because the practice is far too varied and you can cover only one aspect in a yet detailed review, which leads to a conclusion, which is only as good as the way allows it. But awakened by the article of colleagues and in memory of our own attempts with a more complex benchmark course also we are think nevertheless once again on this subject, which will become more explosiveness in particular to the presentation of the first dual-core CPUs.

A fact seems far away from the many speculation however already sure to be: should change the time window between providing the test sample and case of the embargo not for the benefit of the editors and not abate the flood of monthly announcements, many of the considerations and approaches in practice are likely to apply difficult. If this is in the interest of the manufacturer, may – be doubted just with a view to the promotional and prestigious dual-core CPUs-. In other areas, it seems since welcomed per her. Solutions are can be found with certainty only in cooperation. A first approach or an extension of the traditional approach have made the HardTecs4U, but also stress that it is impossible to achieve a such effort in the long term. Ultimately all ways whether your lead him probably own definition to success and none will embody the perfect practical test. The question is only whether you can herausfischen the best way or draw only by referral to all paths at the end of a complex picture can be.

In conclusion, that we do not revise our conclusions for the reasons above at this point taken in the last few months and want to throw on the pile. We are however aware that every conclusion only in his niche had full validity.